Speckle after 5k spp
Speckle after 5k spp
Hello
I have this test scene for testing performance of rendering modes or lighting.
The setup is as; hdr lighting exterior, meshlight interior, thickness glass, glass acceleration on. Render mode is Bidir MLT (From limited tests I found it was the best performing at this type scene).
The render has gone to 5000 s/p after 1.5 hours.
-The main problem is the speckle on the red ball, I assume this is light reflecting off the window from the interior
-Also the clarity of the glass reflection is less than desired (currently the reflected sky is a bit patchy instead of incredibly shiny/slick). Probably related to general performance of HDR lighting.
Does anyone know more tricks or another method to improve upon these issues?
I have this test scene for testing performance of rendering modes or lighting.
The setup is as; hdr lighting exterior, meshlight interior, thickness glass, glass acceleration on. Render mode is Bidir MLT (From limited tests I found it was the best performing at this type scene).
The render has gone to 5000 s/p after 1.5 hours.
-The main problem is the speckle on the red ball, I assume this is light reflecting off the window from the interior
-Also the clarity of the glass reflection is less than desired (currently the reflected sky is a bit patchy instead of incredibly shiny/slick). Probably related to general performance of HDR lighting.
Does anyone know more tricks or another method to improve upon these issues?
Re: Speckle after 5k spp
I assume you used Rainhard Tonemapping.
If I'm right then there is a good chance that your mesh light in the room is waaaaay underpowered compared to the HDRI.
Use linear tonemapping to doublecheck that both emitters are equally strong.
If that isn't the case, then Indigo does process the brighter Lightsource (waaaay) more then the weak one,
since it suggest it to be more important (for example it wouldn't be effective to use 50% of CPU power to a LED light and 50% to a lamp in a room).
Your scene should also work fine with (biDir)PT I think
If I'm right then there is a good chance that your mesh light in the room is waaaaay underpowered compared to the HDRI.
Use linear tonemapping to doublecheck that both emitters are equally strong.
If that isn't the case, then Indigo does process the brighter Lightsource (waaaay) more then the weak one,
since it suggest it to be more important (for example it wouldn't be effective to use 50% of CPU power to a LED light and 50% to a lamp in a room).
Your scene should also work fine with (biDir)PT I think
polygonmanufaktur.de
Re: Speckle after 5k spp
Hi
Actually my HDR gain exponent is set to *(10^8)...to balance with the meshlight whose emission power is actually lowered all the way to 1 (instead of default 1500). Maybe this is not normal.
I will try more render with plain Bidir.
Quick Linear tonemap (with mlt):
Actually my HDR gain exponent is set to *(10^8)...to balance with the meshlight whose emission power is actually lowered all the way to 1 (instead of default 1500). Maybe this is not normal.
I will try more render with plain Bidir.
Quick Linear tonemap (with mlt):
Re: Speckle after 5k spp
Yes, this is light coming from the indoor light bouncing on the window, lighting the ball, passing the window and reaching the camera, this is a typical difficult path.
No magical solution without altering the scene. Increasing the light source size would help. Otherwise, these scenes have to render very long
Etienne
No magical solution without altering the scene. Increasing the light source size would help. Otherwise, these scenes have to render very long
Etienne
Eclat-Digital Research
http://www.eclat-digital.com
http://www.eclat-digital.com
Re: Speckle after 5k spp
Oh, there is a easy one!galinette wrote:No magical solution without altering the scene. Increasing the light source size would help. Otherwise, these scenes have to render very long
Simply raise the emitter power of the interior light to lets say 500%, the reduce the power in the LightLayer accordingly.
By doing so you set the light to higher emitter power and also to higher "Indigo-CPU-time-importance", that should sample the light better (and other light sources less!).
polygonmanufaktur.de
Re: Speckle after 5k spp
My cheap 2 cents will be the standard; make sure the walls aren't 100% (255) white, double check how the window glass is modeled, and sharing the scene always makes things easier
Re: Speckle after 5k spp
I Already thought about having such a text in my SignatureENSLAVER wrote:My cheap 2 cents will be the standard; make sure the walls aren't 100% (255) white, double check how the window glass is modeled, and sharing the scene always makes things easier
polygonmanufaktur.de
Re: Speckle after 5k spp
The magic number 204 should be there tooZom-B wrote:I Already thought about having such a text in my SignatureENSLAVER wrote:My cheap 2 cents will be the standard; make sure the walls aren't 100% (255) white, double check how the window glass is modeled, and sharing the scene always makes things easier
little gallery... http://unverzagt.biz/cottysgallery/
Re: Speckle after 5k spp
I'm not sure, if this is the problem I mention, even removing completely the envmap (black environment) will not solve this. It's a Specular-Diffuse-Specular path. I may be wrong though...Zom-B wrote:Oh, there is a easy one!galinette wrote:No magical solution without altering the scene. Increasing the light source size would help. Otherwise, these scenes have to render very long
Simply raise the emitter power of the interior light to lets say 500%, the reduce the power in the LightLayer accordingly.
By doing so you set the light to higher emitter power and also to higher "Indigo-CPU-time-importance", that should sample the light better (and other light sources less!).
Eclat-Digital Research
http://www.eclat-digital.com
http://www.eclat-digital.com
Re: Speckle after 5k spp
THanks guys for the tips.
I made some adjustments;
-Change to plain Bidir, seems smoother overall.
-Change lights to IES
-Lights are now 4 downlights, 3cm planes
-Ies wall lights exerior
-walls changed to 65% reflect instead of 80%
galinette suggest to increase the light source to help render these reflections on the ball, I figure do the opposite and hope these paths are never found. Also it's more realistic light setup
Render to 2000 spp, 21 minutes. Much faster but overall I think looks similar or better.
Also of note, the ball is directly under the window, probably not a exactly natural placement of furniture, but a bad case.
I made some adjustments;
-Change to plain Bidir, seems smoother overall.
-Change lights to IES
-Lights are now 4 downlights, 3cm planes
-Ies wall lights exerior
-walls changed to 65% reflect instead of 80%
galinette suggest to increase the light source to help render these reflections on the ball, I figure do the opposite and hope these paths are never found. Also it's more realistic light setup
Render to 2000 spp, 21 minutes. Much faster but overall I think looks similar or better.
Also of note, the ball is directly under the window, probably not a exactly natural placement of furniture, but a bad case.
Who is online
Users browsing this forum: No registered users and 62 guests